An Artificial Neural Network Model for Pediatric Mortality Prediction in Two Tertiary Pediatric Intensive Care Units in South Africa. A Development Study.

Michael A Pienaar,Joseph B Sempa,Nicolaas Luwes,Lincoln J Solomon

doi:10.3389/fped.2022.797080

Abstract

ObjectivesThe performance of mortality prediction models remain a challenge in lower- and middle-income countries. We developed an artificial neural network (ANN) model for the prediction of mortality in two tertiary pediatric intensive care units (PICUs) in South Africa using free to download and use software and commercially available computers. These models were compared to a logistic regression model and a recalibrated version of the Pediatric Index of Mortality 3.DesignThis study used data from a retrospective cohort study to develop an artificial neural model and logistic regression model for mortality prediction. The outcome evaluated was death in PICU.SettingTwo tertiary PICUs in South Africa.Patients2,089 patients up to the age of 13 completed years were included in the study.InterventionsNone.Measurements and Main ResultsThe AUROC was higher for the ANN (0.89) than for the logistic regression model (LR) (0.87) and the recalibrated PIM3 model (0.86). The precision recall curve however favors the ANN over logistic regression and recalibrated PIM3 (AUPRC = 0.6 vs. 0.53 and 0.58, respectively. The slope of the calibration curve was 1.12 for the ANN model (intercept 0.01), 1.09 for the logistic regression model (intercept 0.05) and 1.02 (intercept 0.01) for the recalibrated version of PIM3. The calibration curve was however closer to the diagonal for the ANN model.ConclusionsArtificial neural network models are a feasible method for mortality prediction in lower- and middle-income countries but significant challenges exist. There is a need to conduct research directed toward the acquisition of large, complex data sets, the integration of documented clinical care into clinical research and the promotion of the development of electronic health record systems in lower and middle income settings.

Highlights

Prediction of mortality in the pediatric intensive care unit (PICU) has applications in clinical care
Reliable prognostic estimates may allow providers to inform families of the likely outcome of patients admitted to intensive care units limitations in the performance of current models at an individual level hamper their use in patient care [8]
In their recent evaluation of PIM3 in South Africa, Solomon et al found PIM3 to be poorly calibrated in a multicenter prospective study in South Africa

Summary

Introduction

Prediction of mortality in the pediatric intensive care unit (PICU) has applications in clinical care. It has been proposed that miscalibration indicated by significant Hosmer Lemeshow(HL) p-values are likely due to better or worse standards of care in the evaluated PICU as compared to the development study group [18]. In their recent evaluation of PIM3 in South Africa, Solomon et al found PIM3 to be poorly calibrated in a multicenter prospective study in South Africa. They suggested that significant HL p-values in South Africa may be due to case-mix differences between the studied population and the derivation population [13]. In this study we investigate ANNs as a novel method for this application, and apply the standard method of logistic regression, both to the development of a new model and a recalibration of an existing standard model (PIM3)

Objectives

Results

Discussion

Conclusion